Fix research organization and run-level lineage rendering

docs/ARCHITECTURE.md

@@ -135,6 +135,29 @@ An archived Cell is absent from the current map hierarchy and numeric comparison
 archived claims and decisions remain in the historical decisions list with their
 evidence. Archiving does not delete their source snapshots.
 
+### Research organization review and focused lineage
+
+Recorded Run-to-Run derivations are projected between their owning Cells for
+display. The projection preserves source Run edges, merges duplicate visual
+links, and omits intra-Cell replication edges. It does not assert common ancestry
+for every replicate or alter identities. Missing origins, multiple parents and
+projection cycles remain reviewable; no lineage is inferred from labels or dates.
+
+A project-level module may name a stable `attrs.current_focus` node. The default
+canvas then shows that line and its recorded origins in one lane. All research
+tracks and Outline retain access to the complete history. Lane labels preserve
+the recorded research topic even when every condition is a root.
+A Run focus resolves to its owning Cell. Archived focus declarations are ignored;
+unsupported focus targets are reported without hiding the overview. Hierarchy
+ancestors are retained when they provide the recorded lineage fallback.
+
+`organization.py` provides read-only structural hints in `context`, `validate`
+and presentation data. A large condition list without synthesis or comparison,
+wide peer sections and invalid focus references request review. Independent
+study designs may legitimately have no lineage. These hints neither invalidate
+schema nor certify scientific meaning; the host still authors the research
+reasoning separately from implementation flow and the run ledger.
+
 ## Provider boundary
 
 The optional `autoresearch.py` workflow and `backends/shinka.py` adapter execute

research_harness/cli.py

@@ -11,6 +11,7 @@
 from .store import Store, HarnessError, atomic_write, read_json, json_write, validate_state
 from .scan import sync
 from .reasoning import compare, assess, drift
+from .organization import organization
 from .server import export_html, export_site, serve
 
 
@@ -222,7 +223,7 @@ def main(argv: list[str] | None = None) -> int:
                 top = [n for n in nodes if n["kind"] not in {"run", "evaluation", "artifact"}]
                 emit({"project": state["project"], "base_revision": state["revision"], "nodes": top,
                       "node_count": len(nodes), "file_count": len(state["files"]), "drift": drift(state),
-                      "scan": state["scan"], "history": state["history"][-12:],
+                      "scan": state["scan"], "history": state["history"][-12:], "organization": organization(state),
                       "next": "Use files for paginated evidence, context --node ID, or --full. Read evidence before applying a semantic patch."})
         elif args.command == "files":
             state = store.load()
@@ -253,7 +254,7 @@ def main(argv: list[str] | None = None) -> int:
                     except HarnessError as exc:
                         failures.append({"node": n["id"], "error": str(exc)})
             emit({"valid": not failures, "revision": state["revision"], "nodes": len(state["nodes"]),
-                  "evidence_errors": failures, "drift": drift(state),
+                  "evidence_errors": failures, "drift": drift(state), "organization": organization(state),
                   "scope": "Mechanical validation only; not validation of scientific truth or host-model reliability."})
             return 1 if failures else 0
         elif args.command == "export":

research_harness/organization.py

@@ -0,0 +1,70 @@
+"""Read-only structural review hints; not a scientific-quality certificate."""
+from __future__ import annotations
+
+
+def organization(state: dict) -> dict:
+    nodes = {key: n for key, n in state['nodes'].items() if n.get('status') != 'archived'}
+    warnings = []
+    children = {}
+    for n in nodes.values():
+        children.setdefault(n.get('parent_id'), []).append(n)
+
+    def ancestor(node_id, kinds):
+        seen = set()
+        while node_id in nodes and node_id not in seen:
+            seen.add(node_id)
+            n = nodes[node_id]
+            if n['kind'] in kinds:
+                return node_id
+            node_id = n.get('parent_id')
+        return None
+
+    studies = {}
+    for n in nodes.values():
+        if n['kind'] == 'cell':
+            owner = ancestor(n.get('parent_id'), {'module', 'question'})
+            studies.setdefault(owner, []).append(n['id'])
+    related = set()
+    run_lineage = 0
+    for e in state.get('edges', {}).values():
+        if e['relation'] not in {'derived_from', 'compares', 'inspired_by'}:
+            continue
+        source, target = nodes.get(e['source']), nodes.get(e['target'])
+        if not source or not target or source['kind'] not in {'cell', 'run'} or target['kind'] not in {'cell', 'run'}:
+            continue
+        a, b = ancestor(source['id'], {'cell'}), ancestor(target['id'], {'cell'})
+        if a and b and a != b:
+            related.update((a, b))
+            if e['relation'] == 'derived_from' and (source['kind'] == 'run' or target['kind'] == 'run'):
+                run_lineage += 1
+    for owner, cells in studies.items():
+        if len(cells) < 5 or nodes.get(owner, {}).get('attrs', {}).get('role') == 'evidence_audit':
+            continue
+        attrs = nodes.get(owner, {}).get('attrs', {})
+        if not attrs.get('study_summary'):
+            warnings.append({'code': 'missing_study_synthesis', 'node_id': owner, 'cell_ids': cells,
+                             'message': 'Several conditions have no study synthesis. Record the question, comparison, outcome and remaining gap on their study.'})
+        if not any(c in related for c in cells) and not attrs.get('design'):
+            warnings.append({'code': 'unexplained_condition_list', 'node_id': owner, 'cell_ids': cells,
+                             'message': 'Conditions have neither recorded comparisons/lineage nor an explicit independent study design. Review evidence; never invent edges to make a tree.'})
+    for parent, items in children.items():
+        modules = [n for n in items if n['kind'] in {'module', 'question'} and n.get('attrs', {}).get('role') != 'evidence_audit']
+        if len(modules) >= 8:
+            warnings.append({'code': 'wide_research_outline', 'node_id': parent,
+                             'related_ids': [n['id'] for n in modules],
+                             'message': 'Many peer research sections. Review whether they form method families, comparisons, or historical diagnostics; retain independent topics when justified.'})
+    focuses = []
+    for n in nodes.values():
+        focus = n.get('attrs', {}).get('current_focus')
+        if focus is None:
+            continue
+        target = nodes.get(focus) if isinstance(focus, str) else None
+        if (not target or target['kind'] not in {'module', 'question', 'cell', 'run'}
+                or (target['kind'] == 'run' and ancestor(focus, {'cell'}) is None)):
+            warnings.append({'code': 'invalid_current_focus', 'node_id': n['id'],
+                             'message': 'Current focus must reference a visible module, question, cell, or run with a visible owning cell.'})
+        else:
+            focuses.append(focus)
+    return {'needs_review': bool(warnings), 'warnings': warnings,
+            'run_lineage_edges': run_lineage, 'current_focus_ids': sorted(set(focuses)),
+            'scope': 'Structural review hints only. Independent experiments need not form a lineage; schema validity does not certify research organization or scientific conclusions.'}

research_harness/server.py

@@ -11,13 +11,14 @@
 import webbrowser
 from .store import Store, HarnessError, atomic_write, canonical
 from .reasoning import compare, drift
+from .organization import organization
 
 STATIC = Path(__file__).parent / "static"
 
 
 def presentation(store: Store) -> dict:
     state = store.load()
-    return {**state, "comparison": compare(state), "drift": drift(state), "mode": "live"}
+    return {**state, "comparison": compare(state), "drift": drift(state), "organization": organization(state), "mode": "live"}
 
 
 def export_html(store: Store, output: Path, *, include_evidence: bool = False,

research_harness/skill/research-harness/SKILL.md

@@ -80,6 +80,14 @@ below use legacy `rh` as shorthand for that same CLI. Quote all paths.
    from sources, the researcher, or agent interpretation. Do not promote a
    directory hierarchy into this outline. If accounts conflict, retain their
    version and scope instead of choosing a convenient narrative.
+   Keep three separate explanations: research reasoning (failure/question ->
+   hypothesis -> controlled comparison -> result -> decision), the chosen
+   implementation's execution/data flow, and the full run ledger. A batch list
+   or runtime phase list is not a research reasoning chain. Explain why each
+   important result changed the next decision. Reuse the best-supported lineage
+   before opening another branch. When the user selects a main line, record its
+   stable ID as `attrs.current_focus` on the project-level research module;
+   retain parked studies and evidence as history.
 5. **Map the experiment design and testable claims.** Use `rh context WORKSPACE
    --full` or `--node ID` and reuse IDs. Nest modules for project-specific
    research tracks, studies and ablation groups; connect questions → cells → runs
@@ -153,6 +161,12 @@ below use legacy `rh` as shorthand for that same CLI. Quote all paths.
    human-protected corrections. Review source-drift flags without declaring every
    old conclusion invalid.
 9. **Check and visualize.** Run `rh validate WORKSPACE` and `rh compare WORKSPACE`.
+   Inspect `organization` in `context` and `validate`; `valid: true` checks
+   schema/evidence, not whether the research narrative is organized. Resolve or
+   explain structural review hints before delivery. Inspect the rendered map:
+   Run-level derivations must remain visible in the Cell overview. Never infer
+   parentage from version numbers or dates to make a prettier tree. A justified
+   independent comparison does not need artificial lineage edges.
    Current comparisons exclude runs whose metric evidence changed or disappeared;
    their old snapshots remain historical evidence. Address mechanical errors,
    not by weakening provenance. Export with `rh export

research_harness/skill/research-harness/references/PROTOCOL.md

@@ -16,6 +16,29 @@ Relations: `derived_from`, `compares`, `supports`, `contradicts`,
 `uses_checkpoint`, `supersedes`, `inspired_by`. Never confuse method derivation
 with inherited weights. Every edge endpoint must exist.
 
+## Research Organization and Display
+
+The map projects recorded Run-to-Run derivations between their owning Cells.
+This is read-only provenance aggregation, not a claim that every replicate has
+the same ancestry. Duplicate visual links retain their Run sources; intra-Cell
+replication is not a new condition. Multiple origins and cycles remain visible.
+Record the actual Run relationship instead of inventing a Cell-level one for
+the renderer.
+
+An optional `attrs.current_focus: "STABLE_NODE_ID"` on a project-level module
+selects the current research line. The focused view retains recorded origins;
+All research tracks and Outline expose the full history. It does not change
+scientific status, evidence, or run counts.
+Use a visible module, question or Cell ID; a Run ID resolves to its owning Cell.
+Archived declarations and unsupported focus targets do not hide the full map.
+
+Keep the research reasoning chain separate from implementation execution flow
+and the run ledger. Each major step identifies the question, baseline, change,
+observed result and ensuing decision. Source names and runtime phases are not
+research reasoning. The `organization` report in `context`/`validate` provides
+structural review hints; warnings neither invalidate schema nor establish
+scientific quality. Independent studies may legitimately have no lineage edges.
+
 ## Per-version change annotations
 
 Every version, attempt or candidate shown in a research map needs a readable